Biomedical Engineering Reference
In-Depth Information
1
H
N
,
13
C9,
13
C
a
,
13
C
b
and
15
N shifts of 0.62, 1.92, 1.45, 1.31 and 2.89 ppm,
whereas optimal conformational space sampling of an accelerated MD run
that simulated millisecond timescale motions yielded significantly reduced
RMS errors of 0.44, 1.62, 1.01, 1.01 and 1.84 ppm, respectively.
122
The greatest
improvements were observed for regions of the protein that experienced
substantial local motions. The flipside of this time- and ensemble-averaging
methodology can be exploited by using experimental chemical shifts to select
representative structures for ensembles that can reproduce the properties of
intrinsically disordered and partially folded proteins.
123,124
In addition to backbone chemical shifts, the SHIFTX2 program returns
predictions for side-chain
13
C and
1
H nuclei, with overall RMS errors of 0.98
and 0.17 ppm, respectively.
111
The CamShift structure-based approach has
also been extended to selected side-chain sites, with the CH3Shift server
yielding RMS errors for
1
H shifts ranging from 0.13 to 0.20 ppm for Ala, Thr,
Val, Leu and Ile methyl groups
125
and the ArShift server producing errors in
the range 0.17 to 0.26 ppm for Phe and Tyr aromatic
1
H nuclei.
126
These
methods are currently limited by a lack of recorded stereospecific assignments
for methyl groups and because the effects of burial, solvent exposure and side-
chain dynamics are not sufficiently well understood.
125
Main-chain and side-
chain shift predictions are already being applied to validate protein structure
models and MD trajectories,
127
offering both global indicators of quality, such
as ensemble averaged
13
C
a
RMS errors,
119
as well as more focused local
analysis of side-chain conformations.
125,126
CamShift predictions rely on differentiable functions of atomic coordinates
and can be calculated rapidly, so they are well suited to computing forces that
minimise differences between experimental and predicted chemical shiftsin
Monte Carlo and MD simulations.
128,129
Proof of this principle has been
demonstrated by using chemical shift prediction to guide the folding of two
mini-protein (,50 residue) model systems starting from an extended
polypeptide chain, but the method was judged to require too much processor
time for routine use.
128
However, when combined with additional information
in the form of residual dipolar coupling data, the chemical shift MD approach
is capable of determining structures that can only be defined using sparse
restraints, such as the transiently populated excited state of an SH3 domain.
129
A good deal of information about interactions between polar groups in
proteins is encoded in the pH-dependence of chemical shifts, but at present
interpretation of these details relies on the availability of structural data. By
referring to high resolution crystal structures determined under several
different conditions, the Williamson group has performed a comprehensive
analysis of chemical shift changes in the streptococcal GB1 domain between
pH 2 and 10.
130-132
This small, highly stable construct contains a small
selection of titrating groups, including the N- and C-termini, six aspartate, five
glutamate, six lysine side-chains and an N-terminal His-tag, each with K
a
ionisation
1
H,
13
C
15
N
constants characterised
by
NMR
studies
of
and
chemical shifts.
130,132,133
Titration of the carboxylic acid side-chains produced
